BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Phiala Shanahan (MIT) DTSTART:20210323T170000Z DTEND:20210323T180000Z DTSTAMP:20260423T022623Z UID:nhetc/7 DESCRIPTION:Title: Ma chine learning for physics: gauge-equivariant architectures\nby Phiala Shanahan (MIT) as part of NHETC Seminar\n\n\nAbstract\nAs machine learnin g algorithms continue to enable and accelerate physics calculations in nov el ways\, the development of tailored physics-informed machine learning ap proaches is becoming more sophisticated\, impactful\, and important. I wil l give some broad context for this developing area\, with a focus on the c hallenge of exact sampling from known probability distributions as relevan t to lattice quantum field theory calculations in particle and nuclear phy sics. I will discuss in particular flow-based generative models\, and desc ribe how guarantees of exactness and the incorporation of complex symmetri es (e.g.\, gauge symmetry) into model architectures can be achieved. I wil l show the results of proof-of-principle studies that demonstrate that sam pling from generative models can be orders of magnitude more efficient tha n traditional Hamiltonian/hybrid Monte Carlo approaches in this context.\n LOCATION:https://researchseminars.org/talk/nhetc/7/ END:VEVENT END:VCALENDAR